MIXED TREATMENT COMPARISONS TO EVALUATE THE EFFECTIVENESS OF STRATEGIES FOR PREVENTING FIRE RELATED INJURIES IN CHILDREN WITHIN THE HOME Pedro Saramago Goncalves PhD Student, Centre for Health Economics, University of York In collaboration with Professor Alex Sutton, Dr Nicola Cooper, Paula Dhiman: University of Leicester & Professor Denise Kendrick: University of Nottingham. On behalf of the ‘Keeping Children Safe’ project. OBJECTIVES • To identify the most effective (i.e. “best”) strategy for increasing the ownership of fire- related safety equipment in households (smoke alarms, fire extinguishers, fire guards, safe storage of matches/lighters, fire escape plans). METHODS: Mixed Treatment Comparisons • MTC methods (a generalisation of meta-analysis methods) allow comparisons of strategies not directly assessed within any individual primary study. A Trials 3,4 Trials 1,2 B No trials C • MTC uses the totality of the evidence to define a network of evidence STAGES IN MTC 1 •Find the evidence 2 •Define the network 3 •Synthesise the evidence 1. FINDING THE EVIDENCE • Updated Cochrane systematic review of home safety education +/- safety equipment for the prevention of home injuries in childhood and meta-analyses of studies comparing intervention to usual care (Kendrick et al 2007) • Systematic review of reviews with data extraction from relevant primary studies not in meta- analyses in Cochrane review (e.g. smoke alarm education vs. smoke alarm education + free fitted alarm) 2. DEFINING THE NETWORK 5. Education + low 1 trial 2. Education cost/free equipment + fitting 1 trial E.g. 1 trial 2 trials Interventions for increasing 3. Education + low 2 trials ownership of 1. Usual care cost/free equipment functioning smoke alarms 1 trial 3 trials 4. Education + low 3 trials 6. Education + home cost/free equipment inspection + home inspection 3. SYNTHESISING THE EVIDENCE • Mixed treatment random effects model with a binary outcome (have functioning equipment / do not have functioning equipment) • Adjusted for clustering in cluster trials that did not in original analysis (ICCs estimated from individual studies/literature) • Outcomes: – Odds ratio of each intervention compared to one another – Probability of each intervention being the best RESULTS: Probability intervention is the “best” Fire Matches/lighters Fire escape Probability Smoke alarms Fire guards extinguishers safe storage plans Usual care 0.000 0.017 0.008 0.017 0.030 Education 0.000 0.126 0.025 0.144 Education + equipment 0.210 0.630 Education + equipment 0.063 0.477 (not related) Education + equipment + 0.160 0.752 0.126 0.038 home inspection Education + equipment (not related) + home 0.107 inspection Education + equipment + 0.110 home inspection + fitting Education + equipment + 0.120 0.920 fitting Education + home 0.510 0.030 inspection Community campaigns 0.031 0.349 RESULTS: Odds Ratios versus standard care OR versus Matches/lighters Smoke alarms Fire extinguishers Fire guards Fire escape plans standard care (safe storage) Education 1.049 (0.14 to 3.944) 3.136 (0.081 to 15.62) 3.804 (0.09 to 14.2) 4.66 (0.163 to 28.34) Education + equipment 4.901 (0.821 to 16.57) 10.84 (0.326 to 64.6) Education + equipment 1.773 (0.081 to 8.977) 17.82 (0.093 to 94.07) (not related) Education + equipment + 5.862 (1.182 to 20.5) 11.06 (0.465 to 66.94) 2.403 (0.092 to 14.25) 1.244 (0.225 to 4.175) home inspection Education + equipment (not related) + home 2.525 (0.036 to 19.09) inspection Education + equipment + 3.449 (0.537 to 12.24) 2.795 (0.061 to 13.72) 4127 (0.579 to 9299) fitting Education + home 14.21 (0.82 to 73.89) 1.215 (0.07 to 5.706) inspection Community campaigns 1.097 (0.024 to 5.262) 5.268 (0.167 to 25.97) DISCUSSION • Mixed Treatment Comparison Methods allow a coherent analysis of the totality of the evidence • Major assumption of the approach: – The intervention effect estimated by the BC trials would be the same as the intervention effect estimated by the AC and AB trials if they had included B and C arms etc. – Violated if factors interact with intervention effects and these factors imbalanced across interventions (e.g. social group of study participants etc.) • Quality/validity of studies ignored – As for standard meta-analysis, difficult to incorporate • Small number of studies for most comparisons • Results from these analyses are to be used in decision modelling to ascertain cost-effectiveness of the different interventions CONCLUSIONS • Results from the MTC show that more intensive interventions are generally estimated to be more effective (i.e. with the best strategies usually offering free or low cost equipment) • The best strategies varies across equipment types – may partly be due to limitations in the data • Further studies needed to increase precision of estimates and distinguish most effective components of interventions References and contact details • References: Kendrick, D., Coupland, C., Mulvaney, C., Simpson, J. Smith, S.J., Sutton, A., Watson, M., Woods, A. Home safety education and provision of safety equipment for injury prevention. Cochrane Database of Systematic Reviews 2007, CD005014. Caldwell DM, Gibb DM, Ades AE. Validity of indirect comparisons in meta-analysis. The Lancet. 2007;369:270-1. Sutton, A.J., Ades, A.E., Cooper, N.J., Abrams, K.R. Use of indirect and mixed treatment comparisons for technology assessment. Pharmacoeconomics 2008: 26:753-767. • Contact details: email@example.com This study was funded by the National Institute for Health Research (NIHR) under its Programme Grants for Applied Research funding scheme (RP-PG-0407-10231).The views expressed are those of the author(s) and not necessarily those of the NHS, the NIHR or the Department of Health.